语音信号处理中双门限端点检测算法的改进  

An Improved Algorithm of Double Threshold Endpoint Detection Method in Speech Signal Processing

作　　者：黄洋 赵风海[1] 卢景 Huang Yang;Zhao Fenghai;Lu Jing(Tianjin Key Laboratory of Photoelectric Sensor and Sensor Network Technology,College of Electronic Information and Optical Engineering Nankai University,Tianjin 300350,China)

机构地区：[1]南开大学电子信息与光学工程学院,天津市光电传感器与传感网络技术重点实验室,天津300350

出　　处：《南开大学学报（自然科学版）》2021年第2期58-62,共5页Acta Scientiarum Naturalium Universitatis Nankaiensis

摘　　要：语音信号识别系统预处理阶段中一个关键步骤是语音信号的端点检测,其精准性直接决定了整个语音识别系统的识别效果.传统的短时能量和短时过零率双门限检测法中后端点检测存在偏差,且在有噪声的情况下鲁棒性较差.从动态阈值、短时平均过零率、端点修复、动态检测等方面入手,改进了双门限检测法.优化了的端点检测算法,使得语音识别系统能够更精确地识别和提取语音,减少了语音识别时间,提高了识别率.进一步将无用信号和语音信号完全分离开来,有利于后续语音识别的研究.As a key technology in the preprocessing stage of speech recognition system,the accuracy of speech signal endpoint detection determines the performance of speech recognition system.There is deviation in the traditional double threshold endpoint detection method based on short-time energy and short-time zero crossing rate,in addition,poor robustness in the case of noise.Then double threshold detection method was improved from the three aspects of dynamic threshold,endpoint repair,dynamic detection.The optimized endpoint detection algorithm makes the speech recognition system have high accurate recognition and extraction of speech,reducing the time of speech recognition,and improving the recognition rate.Furthermore,the separation of useless signals and speech signals is beneficial to the subsequent speech recognition.

关 键 词：语音信号处理 语音识别 端点检测 双门限 过零率 

分 类 号：TN912.3[电子电信—通信与信息系统]